EVOLUTIONARY ECOLOGY NATURAL HISTORY

Hoverflies are imperfect mimics of wasp colouration

Christopher H. Taylor1 • Tom Reader1 • Francis Gilbert1

Received: 18 June 2015 / Accepted: 1 March 2016 / Published online: 17 March 2016� Springer International Publishing Switzerland 2016

Abstract Many Batesian mimics are considered to be inaccurate copies of their models,

including a number of hoverfly species which appear to be poor mimics of bees and wasps.

This inaccuracy is surprising since more similar mimics are expected to deceive predators

more frequently and therefore have greater survival. One suggested explanation is that

mimics which appear inaccurate to human eyes may be perceived differently by birds, the

probable agents of selection. For example, if patterns contain an ultra-violet (UV) com-

ponent, this would be visible to birds but overlooked by humans. So far, indirect com-

parisons have been made using human and bird responses to mimetic stimuli, but direct

colour measurements of mimetic hoverflies are lacking. We took spectral readings from a

wide range of hoverfly and wasp patterns. They show very low reflectance in the UV range,

and do not display any human-invisible colour boundaries. We modelled how the recorded

spectra would be perceived by both birds and humans. While colour differences between

wasps and hoverflies are slightly more distinct according to human visual abilities, bird

vision is capable of discriminating the two taxa in almost all cases. We discuss a number of

factors that might make the discrimination task more challenging for a predator in the field,

which could explain the apparent lack of selection for accurate colour mimicry.

Keywords Spectrophotometry � Colour analysis � Visual model � Just NoticeableDifference � Batesian mimicry � Syrphidae

Electronic supplementary material The online version of this article (doi:10.1007/s10682-016-9824-9)contains supplementary material, which is available to authorized users.

& Christopher H. Taylorchristopher.taylor@nottingham.ac.uk

1 School of Life Sciences, University of Nottingham, University Park, Nottingham NG7 2RD, UK

123

Evol Ecol (2016) 30:567–581DOI 10.1007/s10682-016-9824-9

Introduction

Colour is widely used by animals as a signal, for example to attract mates (Andersson

1994) or as an anti-predator warning display (Ruxton et al. 2004). However, colour is an

experience as much as a physical property, and therefore the perceived signal depends on

the visual and cognitive abilities of the observer (Endler 1990). For example, male blue tits

(Cyanistes caeruleus) use an ultra-violet (UV) signal to attract a mate, which is striking to

female conspecifics but invisible to humans (Andersson et al. 1998). The butterfly Heli-

conius numata displays a colourful wing pattern that conveys different signals to other

butterflies and to potential predators (Llaurens et al. 2014). Cases like these demonstrate

the importance of considering the signal receiver when assessing the colour component of

any biological signal, and show that doing so can shed new light on well-studied systems.

Some harmless organisms attempt to deceive predators by mimicking the display of a

more dangerous ‘‘model’’, in a process known as Batesian mimicry (Bates 1862). Mimetic

displays can incorporate a range of different cues, including shape (Jones et al. 2013),

pattern (Bain et al. 2007) and movement (Golding et al. 2005), but among these, colour is

thought to be particularly salient to predators (Marples et al. 1994; Aronsson and Gam-

berale-Stille 2012; Kazemi et al. 2014). Most experimental evidence suggests that Batesian

mimics should gain the greatest protection by resembling their models as accurately as

possible (Dittrich et al. 1993; Lindstrom et al. 1997). However, to human eyes there is

great variation in the degree of resemblance between mimics and models in nature, which

raises the question of why the less accurate mimics persist in the face of predicted selection

towards perfect resemblance (Edmunds 2000; Kikuchi and Pfennig 2013).

One proposed solution to the problem is that a perceived lack of mimetic accuracy as

observed by humans might be specific to our particular visual abilities (Cuthill and Bennett

1993). If mimicry is in the ‘‘eye of the beholder’’, those mimics that we (as humans)

consider to be inaccurate might be highly accurate when viewed by an observer with

different sensory and cognitive capabilities. A key part of Cuthill and Bennett’s (1993)

argument was that in systems with avian predators, the birds’ ability to detect UV light

(Chen and Goldsmith 1986) might lead them to interpret patterns very differently to

humans.

In their ‘‘eye of the beholder’’ hypothesis, Cuthill and Bennett (1993) make particular

reference to hoverflies (Diptera: Syrphidae), which, together with their models, are a key

study system for understanding the evolution of imperfect mimicry (see e.g. Dittrich et al.

1993; Azmeh et al. 1998; Holloway et al. 2002; Penney et al. 2012). The family comprises

a large number of species, many of which are abundant and widespread, ranging from non-

mimetic to highly accurate mimics of various hymenopteran models (bees and wasps;

Apidae and Vespidae), with a wide range of accuracy in between (Gilbert 2005; Rotheray

and Gilbert 2011).

Predation from birds is thought to provide the main selective pressure on hoverfly

colour patterns (Waldbauer 1988; Gilbert 2005; Bain et al. 2007). Bees and wasps are

unpalatable to most birds due to their sting and tough cuticle (Mostler 1935), although a

few specialists do prey on them despite these defences, usually by removing the sting prior

to consumption (Birkhead 1974). Hoverflies do not appear to have any chemical or

physical defence, being readily accepted and consumed by birds that have not previously

encountered a model (Mostler 1935). Some models, such as Apis mellifera and Vespula

vulgaris, are widespread in the Palearctic and overlap extensively with the ranges of many

hoverfly species, while other potential models (e.g. Ectemnius cavifrons) are more

568 Evol Ecol (2016) 30:567–581

123

restricted geographically and may be present in only part of the range of a given mimic

(Richards 1980). Most hoverflies, bees and wasps are also likely to overlap over a finer,

microhabitat scale, with many being frequent flower visitors. In the UK, most species of

hoverfly first emerge between March and May and remain active until at least September

(Stubbs and Falk 2002), with workers of social Hymenoptera generally reaching peak

abundance in July/August (Richards 1980).

Given the discrepancy between bird and human visual abilities, it is vital to consider

avian perception of the mimetic signals (Cuthill and Bennett 1993). Suitable methods are

well-developed for both collection of spectral data and its subsequent interpretation

through the eyes of a particular observer (Endler 1990; Vorobyev and Osorio 1998; Endler

and Mielke 2005). These methods have been used to investigate mimetic accuracy in

animals such as fish (Cheney and Marshall 2009), butterflies (Llaurens et al. 2014) and

salamanders (Kraemer and Adams 2014), but to our knowledge, detailed colour analysis is

lacking for hoverflies and their hymenopteran models.

Most animals, including birds, are thought to perceive the chromatic (hue and satura-

tion) and achromatic (brightness) components of colour separately, and the information in

these different channels may be used in different contexts by the signal receiver (Giurfa

et al. 1997; Osorio et al. 1999). Chromatic stimuli are useful for comparison among

disparate objects, as the chromatic properties do not change much under different illu-

mination conditions. Achromatic signals are strongly affected by illumination, but are

useful for detecting local changes in spectral properties, such as at the border between two

colour patches (Osorio et al. 1999). From this, we predict two possible ways in which

spectral properties could be used by an observer to discriminate between models and

mimics. The absolute values of chromatic stimuli could be important, as birds have been

shown to learn and recognise particular colour combinations in potential prey (Svadova

et al. 2009; Aronsson and Gamberale-Stille 2012; Kazemi et al. 2014). For achromatic

stimuli, the absolute values are less likely to be relevant, since they are difficult to compare

for samples that are separated in space and time (Osorio et al. 1999). However, the

achromatic contrast between colour patches within a single pattern will be easier to detect,

and could form an important signal (Aronsson and Gamberale-Stille 2013).

In this study, we present data on both chromatic and achromatic components of the

colours of wasp-mimicking hoverflies and their potential models, confirming that there is

no ‘‘hidden’’ signal in the patterns of either taxon. Then, we interpret the colours through

the eyes of avian predators and estimate the level of mimetic accuracy that is achieved. In

doing so, we show that mimicry of wasp colours by hoverflies is, to varying degrees,

imperfect.

Materials and methods

Specimens

Insects were collected using a hand net from wild communities in Nottinghamshire, UK

and surrounding areas, during July to September 2014. Target insects were any hoverflies

or stinging Hymenoptera bearing a two-colour (typically yellow and black) pattern

(Fig. 1), but excluding bumblebees and their putative mimics because they are very likely

part of a different mimicry ring (Gilbert 2005), and their hairiness makes taking reliable

Evol Ecol (2016) 30:567–581 569

123

colour measurements difficult. A total of 247 individuals were identified to species level

and sexed using relevant keys (Richards 1980; Stubbs and Falk 2002).

Specimens were euthanised by freezing for 10–20 min on the day of capture and then

pinned. Colour measurements (see below) were taken within 1 h of death to minimise any

colour changes that might occur (colours of some species fade during the days following

death: C Taylor, pers. obs.).

Eight different model species were sampled, but only four were found more than twice:

Apis mellifera (N = 14), Vespula vulgaris (N = 10), V. germanica (N = 3) and Vespa

crabro (N = 5). We know from both theory (Getty 1985) and experiments (Lindstrom

et al. 1997) that a model’s importance in shaping predator behaviour increases with its

abundance, and therefore we have excluded rare models (N\ 3) from the bulk of the

analysis. However, for comparison, we also conducted a repeat analysis using all eight

model species.

Spectrophotometry

Reflectance measurements were taken using a 100 lm bifurcating optic fibre probe (Ocean

Optics, Dunedin, FL, USA, custom spec) with one fibre connected to a pulsed xenon light

source (Ocean Optics PX-2) and the other to a spectrophotometer (Ocean Optics USB

2000? UV–VIS–ES). The probe was held steady and targeted using a micro-manipulator

(Prior, Cambridge, UK). The probe was fixed at an angle of 45� to horizontal, and the patchunder measurement was placed as close to horizontal as possible. A custom-made alu-

minium probe cover cut off at an angle of 45� aided with this alignment, and also helped to

maintain a constant distance (approximately 2 mm) between the specimen and the probe

(Endler 1990). The light source pulsed at a frequency of 50 Hz and spectral readings were

integrated over 10 pulses, or 200 ms. Measurements were recorded for wavelengths over

the range 300–700 nm at intervals of 0.4 nm. Measurements were taken in relation to a

white standard (Ocean Optics WS-1 Diffuse Reflectance Standard) and recalibrated to the

standard after approximately every four specimens in order to account for lamp drift. All

Fig. 1 Examples of colour and pattern variation in hymenopteran (a, b) and hoverfly (c–j) abdominalpatterns. Scale bars each show 1 mm. a Vesupla vulgaris. b Apis mellifera. c Eristalis tenax. d Eristalispertinax. e Melangyna labiatarum. f Sericomyia silentis. g Syrphus ribesii. h Sphaerophoria scripta.i Episyrphus balteatus. j Platycheirus albimanus. (Color figure online)

570 Evol Ecol (2016) 30:567–581

123

measurements were taken in a dark room with the xenon lamp being the only source of

light.

Pilot testing revealed that readings taken from within 0.2 mm of a colour border were

inaccurate (see Electronic Supplementary Material) and we therefore targeted the centres

of insect colour patches that were at least 0.5 mm in width. In order to check that, in doing

so, we did not overlook any colour boundaries invisible to humans, we moved the probe

across adjacent areas and monitored any changes to the spectral read-out in real time. We

recorded an example of such a process in the form of a transect along the abdomen of an

individual of Helophilus hybridus, with intervals of 0.2 mm.

For each hoverfly or wasp specimen, we took spectral readings from both ‘black’ (low

reflectance, black or dark red/brown to human eyes) and ‘coloured’ (higher reflectance,

usually yellow or orange to human eyes) patches of the abdomen where possible. In a few

cases, patches of one type were too small to take accurate readings and therefore we only

recorded spectra of the predominant patch type in those cases. At least three readings were

taken for both of the patch types (where present), ideally taken from different patches on

different abdominal tergites, again limited in cases where patches were small or absent on

some tergites.

Spectral analysis

Analysis was carried out in R version 3.1.2 (R Core Team 2014) making use of the package

‘pavo’ for spectral processing and visual models (Maia et al. 2013). Noise was removed

from the spectra using loess smoothing over a span of 0.4. Any smoothed spectra showing

negative reflectance values, which can occasionally result from noise or a drift in cali-

bration, were adjusted by adding a constant to the spectrum such that the minimum

reflectance value was zero. ‘‘Brightness’’ was calculated as the mean reflectance value

across the whole spectrum (300–700 nm).

Given that the main selective pressure on hoverfly mimicry is considered to come from

passerine birds (Dlusski 1984; Gilbert 2005) but that no single species stands out as an

obvious candidate, we modelled the colour perception on a generalised ‘UV-type’ retina,

with four cone types (U, S, M and L) with peak sensitivity at 372, 456, 544 and 609 nm

respectively (Odeen and Hastad 2003; Maia et al. 2013). Achromatic stimulation was

based on a blue tit double cone with peak sensitivity of 566 nm. We used models of photon

catch to calculate cone stimulation values for each spectrum (Vorobyev and Osorio 1998;

Maia et al. 2013). Given the wide range of species included in this study, it is not possible

to record precise illumination conditions that will be valid for all sampled individuals;

however all were collected from locations with low canopy cover, and therefore we

modelled illumination as ‘D65’, which is representative of daylight in open areas (Endler

1993). We then used receptor noise models to calculate the chromatic (DS) or achromatic

(DL) contrast between a given pair of spectra, with units of ‘Just Noticeable Differences’ orJNDs (Vorobyev and Osorio 1998; Maia et al. 2013), and based on a Weber fraction (a

measure of signal to noise ratio) of 0.06 (Olsson et al. 2015).

In accordance with the way in which birds are thought to perceive spectral information

(Osorio et al. 1999), we analysed chromatic and achromatic components of the signals

separately. To compare the chromatic properties of patterns from a pair of species, we

treated black and coloured patch types separately, and calculated DS for each. We then

made the assumption that, in attempting to discriminate two patterns, a predator will attend

to the patch type within the patterns that shows the larger difference. Thus the chromatic

Evol Ecol (2016) 30:567–581 571

123

distance within a given model-mimic pairing was taken as the larger of the DS values for

black and coloured patches.

As achromatic signals are typically used to detect within-pattern variation (Osorio et al.

1999), we calculated the within-pattern achromatic contrast (DL) between the black and

coloured patches for each individual insect. We then calculated the absolute difference

between model and mimic in values of within-pattern contrast as a measure of achromatic

distance.

We repeated the model-mimic comparisons using a different visual model based on

human vision, in order to examine any differences from bird perception. Human cone

sensitivity data was taken from Stockman and Sharpe (2000) and we assumed a Weber

fraction of 0.018 (Wyszecki and Stiles 2000). Achromatic stimulation was calculated as the

sum of M and L cones (Wyszecki and Stiles 2000). To compare achromatic perception

between the two systems, we regressed human against avian estimates of within-pattern

contrast across the insect species sampled, with the intercept fixed at zero. The slope value

gives an estimate for the ratio in achromatic sensitivity between birds and human. We

carried out similar regressions on chromatic contrast data, with separate regressions for the

four different model species and for the two patch types (these eight sets of data could not

be pooled as they are not independent of each other).

Results

We examined spectra from 209 individual hoverflies of 33 species, and 38 individual

Hymenoptera of eight species, sampling both ‘‘black’’ (very low reflectance) and

‘‘coloured’’ (higher reflectance; usually yellow or orange) patch types within the pattern.

At no point did we detect a marked change in spectral properties of any individual that did

not correspond to a human-visible boundary (see example with Helophilus pendulus,

Fig. 2). None of the patterns sampled has a strong UV component in either the coloured or

black patches (Fig. 3).

Human ‘‘Just Noticeable Difference’’ (JND) estimates for within-pattern achromatic

contrast (DL) are related to but considerably higher than the avian equivalents

(slope = 4.4, r2 = 0.996, p\ 0.001). Human and avian JND estimates for chromatic

similarity (DS) are closer to each other, but human values are usually higher. Slopes for

black patches range from 0.88 to 1.41, and for coloured patches from 1.19 to 1.94 (all

r2[ 0.75, p\ 0.001; Fig. 4). Hence, the colour differences that we perceive among

model and mimic species are generally rather larger than those evident to avian

predators.

All remaining values in the results section are calculated with respect to avian vision.

The four main model species (those with N C 3) are distinguishable from each other in

terms of their spectra (Fig. 3). For coloured patches, chromatic contrast ranges from 2 to 12

JNDs (Table S1). Differences among black patches are smaller, ranging from 0.6 to 4.7,

with the largest differences being between Vespa crabro and the other three models. The

three vespid species (Vespa crabro, Vespula vulgaris and V. germanica) have similar levels

of within-pattern achromatic contrast (DL = 40–45 JNDs), whereas contrast for Apis

mellifera is much lower (DL = 22; Table 1).

All of the mimic species sampled are theoretically distinguishable from each of the four

main model species in chromatic terms, although some have DS values only just larger

than one (e.g. Episyrphus balteatus differs from A. mellifera by DS = 1.3; Table 1,

572 Evol Ecol (2016) 30:567–581

123

Figs. S2 and S3). The species sampled are split roughly half and half between being most

similar to A. mellifera (15) and Vespula vulgaris (14), with two being closest to Vespa

crabro and two to Vespula germanica.

Achromatic differences span a wider range of values than chromatic differences, and are

usually larger than the latter (Fig. 5). The hoverflies generally show lower within-pattern

contrast than the Hymenoptera (Fig. S4), but some model-mimic pairings were highly

similar in achromatic terms, with five mimics differing from their closest model by DL\ 1

(Table 1). When mimics are allocated to models according to the lowest achromatic dif-

ference, we find twelve mimics of Vespula vulgaris, ten mimics of A. mellifera, ten of

Vespa crabro and one of Vespula germanica. Agreement between the chromatic and

achromatic measures is poor—the closest model in chromatic terms matches the achro-

matic for only 14 of the 33 mimics (Table 1).

We repeated the above analysis taking into account all eight sampled species of

Hymenoptera, including those with very low abundance. Results in this re-analysis were

very similar, with only nine of 33 hoverfly species having one of these rare species as their

closest model (Table S2).

A B

1 5 10 15 20

12 18 11 19 8

9 10 1 4 3 2 20 17 5 7 15 13 16 22 6 21 14

C

Fig. 2 Colour variation along a transect on an abdomen of Helophilus hybridus. a The abdomen, withlocations of spectral readings shown in red. b All 22 spectra from the transect. c Variation in spectralbrightness along the transect. (Color figure online)

Evol Ecol (2016) 30:567–581 573

123

Ame

Ame

Vvu

Vvu

Vge

Vge

Vcr

Vcr

PalMsc

MscPal

Hpe

Hpe

Mfl

Mfl

Ear

Ear

Epe

Epe

Ete

Ete

Eho

Eho

Vpe

Vpe

Ssi

Ssi

Eba

Eba

Lgl

Lgl

Egr

Egr

Ssc

Ssc

Mci

Mci

Mla

Mla

Sri

Sri

Svi

Svi

Sto

Sto

[Hymenoptera] Bacchini

Eristalini (1) Eristalini (2)

Sericomyini and Volucellini Syrphini (1)

Syrphini (2) Syrphini (3)

0

10

20

30

40

50

0

10

20

30

40

50

0

10

20

30

40

50

0

10

20

30

40

50

300 400 500 600 700 300 400 500 600 700

Wavelength (nm)

Ref

lect

ance

Patch typeBlack

Coloured

Fig. 3 Reflectance spectra for all sampled species with N[ 3. Solid and dashed lines show means for blackand coloured patches respectively, shaded areas show standard error. For species abbreviations, see Table 1

574 Evol Ecol (2016) 30:567–581

123

Discussion

This study represents the first attempt to characterise, in detail, the colours of hoverflies

and their hymenopteran models. From our measurements of insect specimens, we find no

evidence that there are pattern elements or colour boundaries in either hoverflies or their

models that are invisible to the human eye, a fact which until now has only been indirectly

inferred (Green et al. 1999; Penney et al. 2012). Our estimates of colour similarity

according to the sensory abilities of humans and birds correlated closely with one another.

Model and mimic colours were usually less distinct (smaller JND values) when cal-

culated using the avian as opposed to the human visual model. This tallies well with recent

behavioural data, which have shown that humans are at least as good at discriminating

colours as chickens are, thanks to lower levels of receptor noise (Olsson et al. 2015).

Nonetheless, none of the mimics differs from its nearest model by less than one avian JND,

which implies that, in the eyes of birds, any given mimic-model pair should in theory be

distinguishable in terms of colour (Vorobyev and Osorio 1998).

A number of researchers have speculated that a threshold of one JND may not be

realistic in a natural context, instead adopting higher threshold values in the range two to

four (Siddiqi et al. 2004; Feeney et al. 2014; Limeri and Morehouse 2014). There is no

behavioural evidence to support the choice of a particular threshold other than one (Olsson

et al. 2015) but there are a number of factors which might make the discrimination task

more difficult for a predator in the wild than in the controlled laboratory settings on which

Black Coloured

0

10

20

30

40

50

0 10 20 0 10 20Chromatic contrast in avian vision

Chr

omat

ic c

ontra

st in

hum

an v

isio

n

model

Ame

Vcr

Vge

Vvu

Fig. 4 Comparison of estimates of chromatic contrast as calculated in models based on avian and humanvision. Each point represents a single model-mimic pairing. Values are in units of Just NoticeableDifference. The two panels show data on black and coloured patch types separately. For modelabbreviations, see Table 1

Evol Ecol (2016) 30:567–581 575

123

JND values are based. Firstly, predators will rarely have the opportunity to compare two

prey items side by side. Temporal separation of the stimuli, as experienced by a predator

learning to discriminate between models and mimics, will increase the difficulty of the task

(Dyer and Neumeyer 2005).

Secondly, predators may not have the opportunity to view their prey from close range

before deciding whether to attack, reducing the visual information available. It is not clear

at what typical distance a bird might make its decision. Dlusski (1984) showed that

hoverflies would not give flight until a mock predator approached to within 30 cm or less,

suggesting that relatively close inspection might be possible on occasion, but he also

observed that insectivorous birds could find prey from a distance of 10 m or more. The

spatial resolution of the perceived pattern will decline with distance. Passerines can dis-

criminate objects separated by a visual angle of about 1–3 arc min (0.017–0.05�; Donner1951). This would give a spatial resolution of about 0.2 mm at a distance of 30 cm, which

would be enough to perceive the pattern of most hoverflies, but even at a distance of a few

metres, very little detail would be visible (2 mm resolution at a distance of 3 m). At that

range, rather than perceiving separate patches of colour, the bird would perceive a spec-

trum that is an average of the two spectra (weighted by area), but it might still be able to

learn differences among species based on this colour information.

Thirdly, if the prey is moving, this may make the discrimination more difficult. Notably,

motion creates blur which will make borders within the colour pattern less distinct,

0

10

20

30

40

10 20Chromatic distance

Ach

rom

atic

dis

tanc

e

model

Ame

Vcr

Vge

Vvu

Fig. 5 Comparison of chromaticand achromatic distancesbetween mimic species and eachof the four main model species.Each point represents a singlemodel-mimic species pairing. Formodel abbreviations, see Table 1

576 Evol Ecol (2016) 30:567–581

123

Table 1 Achromatic and chromatic distances of each hoverfly species to its closest model

Species Abbrev. Achromatic Chromatic

Internalcontrast

Closestmodel

Dista Closestmodel

Distb Patchcolor

Thoraxwidth

Mimics

Chrysotoxum arcuatum Car 41.2 Vvu 1.2 Vge 9.3 C 2.6

Dasysyrphus albostriatus Dal 55.7 Vcr 10.4 Vvu 8.5 C 2.4

Dasysyrphus tricinctus Dtr 49.3 Vcr 4.0 Vvu 11.4 C 2.5

Epistrophe grossulariae Egr 37.5 Vvu 2.5 Ame 3.8 B 3.2

Episyrphus balteatus Eba 25.9 Ame 4.0 Ame 1.3 C 2.2

Eristalis arbustorum Ear 29.6 Ame 7.7 Ame 3.5 B 3.2

Eristalis horticola Eho 35.5 Vvu 4.5 Vvu 2.9 C 3.5

Eristalis interruptus Eip 32.8 Vvu 7.2 Vcr 5.4 C 3.4

Eristalis pertinax Epe 31.3 Vvu 8.7 Ame 1.6 C 3.7

Eristalis tenax Ete 27.4 Ame 5.6 Vcr 3.6 B 4.4

Eupeodes latifasciatus Ela 52.4 Vcr 7.0 Vge 2.6 C 2.0

Helophilus hybridus Hhy 45.9 Vcr 0.6 Vvu 1.9 C 3.8

Helophilus pendulus Hpe 37.9 Vvu 2.1 Vvu 2.2 B 3.0

Leucozona glaucia Lgl 35.4 Vvu 4.6 Ame 16.2 C 2.6

Melangyna labiatarum Mla 53.5 Vcr 8.2 Vvu 10.7 C 1.8

Melanostoma scalare Msc 27.7 Ame 5.8 Ame 2.6 C 1.6

Meliscaeva auricollis Mau 34.6 Vvu 5.4 Vvu 4.4 C 2.0

Meliscaeva cinctella Mci 30.3 Ame 8.4 Ame 3.4 C 1.9

Myathropa florea Mfl 34.1 Vvu 5.9 Vvu 3.9 C 3.8

Parhelophilus versicolor Pve 43.0 Vge 1.0 Vvu 2.7 C 2.9

Platycheirus albimanus Pal 28.7 Ame 6.8 Ame 10.5 C 1.8

Platycheirus clypeatus Pcl 22.9 Ame 1.0 Ame 2.2 B 1.7

Platycheirus occultus Poc 21.5 Ame 0.3 Ame 13.0 C 1.5

Sericomyia silentis Ssi 63.9 Vcr 18.6 Vvu 3.1 B 4.5

Sphaerophoria scalare Ssc 38.5 Vvu 1.5 Vvu 2.7 C 1.6

Syritta pipiens Spi 22.3 Ame 0.5 Ame 4.5 C 1.6

Syrphus ribesii Sri 51.1 Vcr 5.8 Vvu 4.9 C 2.8

Syrphus torvus Sto 44.9 Vcr 0.5 Vvu 7.3 C 2.9

Syrphus vitripennis Svi 46.8 Vcr 1.5 Vvu 5.5 C 2.4

Volucella inanis Vin 44.9 Vcr 0.4 Ame 5.7 B 4.8

Volucella pellucens Vpe 32.9 Vvu 7.1 Ame 8.2 C 4.9

Volucella zonaria Vzo 38.6 Vvu 1.4 Ame 3.8 B 6.1

Xylota segnis Xse 15.4 Ame 6.5 Ame 5.7 C 2.6

Models

Ancistrocerus trifasciatus Atr 45.3 2.0

Apis mellifera Ame 21.9 3.6

Ectemnius cavifrons Eca 65.7 2.9

Ectemnius continuus Eco 65.9 2.8

Mellinus arvensis Mar 70.3 2.2

Vespa crabro Vcr 45.3 5.5

Evol Ecol (2016) 30:567–581 577

123

although (at least in humans) cognitive processing can remove much of the blur (Burr

1980).Hoverflies are fast fliers and highly active, but one might expect birds to target them

when they are at rest on flowers or vegetation rather than in flight (Dlusski 1984), which

would minimise the impact of blur. Motion blur would have a similar effect to the low

resolutions described above, in that above a certain level, the pattern would not be visible,

and colours of different patches would become merged together. However, a strong colour

difference would certainly still be visible under those circumstances.

Importantly, there is considerable variation among species in the levels of mimetic

accuracy, with several showing chromatic contrast of less than three JNDs with their

nearest model, and others with values of 10 or more. In their natural context, it is likely that

the most accurate hoverflies are more or less ‘‘perfect’’ colour mimics. However, contrary

to the ‘‘eye of the beholder’’ hypothesis (Cuthill and Bennett 1993), those mimics at the

lower end of the accuracy scale should be clearly distinguishable from their models, even if

we allow for the difficulties described above. If birds do indeed provide the main selective

pressure on hoverfly colours, the observed variation in mimetic accuracy cannot be

explained solely by the eye of the beholder hypothesis.

Our data do hint at an alternative explanation for at least some instances of mimetic

inaccuracy. The model species that we sampled were all distinguishable in terms of their

colours; even Vespula vulgaris and V. germanica, two very closely related wasps, differ by

five JNDs. Models and mimics do not segregate neatly according to their colour (Fig-

ures S2 and S3). In contrast to other characteristics such as antenna length, in which

models and mimics are consistently separated (Penney et al. 2012), there is no simple rule

that could be followed to reliably distinguish hoverflies from wasps on the basis of colour.

A predator can therefore adopt one of two strategies: it can learn each species and its

colour entirely separately, which would carry a high cognitive burden as well as requiring

repeated sampling of potentially dangerous prey (Kikuchi and Sherratt 2015), or it may

generalise over a range of colours (Richards-Zawacki et al. 2013; Vesely et al. 2013). In

the latter strategy, a large area of colour space would be protected, including many (though

perhaps not all) of the hoverfly colours (Figs. S2 and S3), and this would explain the lack

of selection for increased mimetic accuracy in those cases.

It is interesting to note that, in the majority of cases, achromatic distances between

model and mimic are larger than chromatic distances (Fig. 5 and Table 1). This implies

that the chromatic properties of the colour pattern may be under stronger selection from

predators than the achromatic properties. The same appears to be true of mimetic

Table 1 continued

Species Abbrev. Achromatic Chromatic

Internalcontrast

Closestmodel

Dista Closestmodel

Distb Patchcolor

Thoraxwidth

Vespula germanica Vge 44.0 3.6

Vespula vulgaris Vvu 40.0 3.1

All values are given in units of Just Noticeable Differencesa Achromatic distances are calculated as the absolute difference between values of internal pattern contrast(that is, the achromatic distance between coloured and black patches) between the model and mimicb Chromatic distances are the DS values between model and mimic for whichever patch type (coloured orblack, indicated in the ‘patch colour’ column) has the larger DS

578 Evol Ecol (2016) 30:567–581

123

salamanders (Kraemer and Adams 2014), another system in which birds are thought to

provide the main selective pressure on colours. Birds may find it difficult to compare

achromatic signals that are separated in time and space given changing light conditions

(Osorio et al. 1999). In addition, their sensitivity to achromatic contrast declines at small

spatial scales (\1� visual angle; Ghim and Hodos 2006), meaning that they may exert little

selection for accurate mimicry of luminance properties.

Numerous experiments have demonstrated that colour is an important stimulus for

predators attempting to discriminate among prey items (Morrell and Turner 1970; Svadova

et al. 2009; Vesely et al. 2013; Kazemi et al. 2014). However, all of these experiments have

used stimuli that are well separated in colour space, corresponding to different named

colour categories. More behavioural studies are needed to establish predator responses to

colour stimuli that differ by more subtle degrees, and to separate the response to achro-

matic and chromatic properties of the stimulus. Most importantly, future work on inac-

curate mimicry must consider the natural history of the predator–prey interactions in order

to take account of highly influential variables such as viewing distance and movement.

Acknowledgments We would like to thank John Endler for advice on spectrophotometry methods, MarkStrickland for manufacture of our custom probe-cover, and two anonymous reviewers for invaluablecomments on our manuscript. This research was partly funded by a small equipment grant from theUniversity of Nottingham.

References

Andersson MB (1994) Sexual selection. Princeton University Press, PrincetonAndersson S, Ornborg J, Andersson M (1998) Ultraviolet sexual dimorphism and assortative mating in blue

tits. Proc R Soc Lond B 265(1395):445–450Aronsson M, Gamberale-Stille G (2012) Colour and pattern similarity in mimicry: evidence for a hierar-

chical discriminative learning of different components. Anim Behav 84(4):881–887Aronsson M, Gamberale-Stille G (2013) Evidence of signaling benefits to contrasting internal color

boundaries in warning coloration. Behav Ecol 24(2):349–354Azmeh S, Owen J, Sørensen K, Grewcock D, Gilbert F (1998) Mimicry profiles are affected by human-

induced habitat changes. Proc R Soc Lond B 265(1412):2285–2290Bain RS, Rashed A, Cowper VJ, Gilbert FS, Sherratt TN (2007) The key mimetic features of hoverflies

through avian eyes. Proc R Soc Lond B 274(1621):1949–1954Bates HW (1862) XXXII. Contributions to an insect fauna of the amazon valley. Lepidoptera: Heliconidæ.

Trans Linn Soc Lond 23(3):495–566Birkhead TR (1974) Predation by birds on social wasps. Br Birds 67(6):221–229Burr D (1980) Motion smear. Nature 284(5752):164–165Chen DM, Goldsmith TH (1986) Four spectral classes of cone in the retinas of birds. J Comp Physiol A

159(4):473–479Cheney KL, Marshall NJ (2009) Mimicry in coral reef fish: how accurate is this deception in terms of color

and luminance? Behav Ecol 20(3):459–468Cuthill IC, Bennett ATD (1993) Mimicry and the eye of the beholder. Proc R Soc Lond B

253(1337):203–204Dittrich W, Gilbert F, Green P, Mcgregor P, Grewcock D (1993) Imperfect mimicry: a pigeon’s perspective.

Proc R Soc Lond B 251(1332):195–200Dlusski G (1984) Are dipteran insects protected by their similarity to stinging Hymenoptera? Byull Mosk

O-Va Ispyt Prir Otd Biol 89:25–40Donner KO (1951) The visual acuity of some passerine birds. Acta Zool Fenn 66:1–40Dyer AG, Neumeyer C (2005) Simultaneous and successive colour discrimination in the honeybee (Apis

mellifera). J Comp Physiol A 191(6):547–557Edmunds M (2000) Why are there good and poor mimics? Biol J Linn Soc 70(3):459–466Endler JA (1990) On the measurement and classification of colour in studies of animal colour patterns. Biol

J Linn Soc 41(4):315–352

Evol Ecol (2016) 30:567–581 579

123

Endler JA (1993) The color of light in forests and its implications. Ecol Monogr 63(1):2–27Endler JA, Mielke PWJ (2005) Comparing entire colour patterns as birds see them. Biol J Linn Soc

86:405–431Feeney WE, Stoddard MC, Kilner RM, Langmore NE (2014) ‘‘Jack-of-all-trades’’ egg mimicry in the brood

parasitic Horsfield’s bronze-cuckoo? Behav Ecol 25(6):1365–1373Getty T (1985) Discriminability and the sigmoid functional response: how optimal foragers could stabilize

model-mimic complexes. Am Nat 125(2):239–256Ghim MM, Hodos W (2006) Spatial contrast sensitivity of birds. J Comp Physiol A 192(5):523–534Gilbert F (2005) The evolution of imperfect mimicry. In: Fellowes M, Holloway G, Rolff J (eds) Insect

evolutionary ecology. CABI, Wallingford, pp 231–288Giurfa M, Vorobyev M, Brandt R, Posner B, Menzel R (1997) Discrimination of coloured stimuli by

honeybees: alternative use of achromatic and chromatic signals. J Comp Physiol A 180(3):235–243Golding YC, Edmunds M, Ennos AR (2005) Flight behaviour during foraging of the social wasp Vespula

vulgaris (Hymenoptera: Vespidae) and four mimetic hoverflies (Diptera: Syrphidae) Sericomyiasilentis, Myathropa florea, Helophilus sp. and Syrphus sp. J Exp Biol 208(23):4523–4527

Green PR, Gentle L, Peake TM, Scudamore RE, McGregor PK, Gilbert F, Dittrich WH (1999) Conditioningpigeons to discriminate naturally lit insect specimens. Behav Processes 46(1):97–102

Holloway G, Gilbert F, Brandt A (2002) The relationship between mimetic imperfection and phenotypicvariation in insect colour patterns. Proc R Soc Lond B 269(1489):411–416

Jones RT, Poul YL, Whibley AC, Merot C, ffrench-Constant RH, Joron M (2013) Wing shape variationassociated with mimicry in butterflies. Evolution 67(8):2323–2334

Kazemi B, Gamberale-Stille G, Tullberg Birgitta S, Leimar O (2014) Stimulus salience as an explanation forimperfect mimicry. Curr Biol 24(9):965–969

Kikuchi DW, Pfennig DW (2013) Imperfect mimicry and the limits of natural selection. Q Rev Biol88(4):297–315

Kikuchi DW, Sherratt TN (2015) Costs of learning and the evolution of mimetic signals. Am Nat186(3):321–332

Kraemer AC, Adams DC (2014) Predator perception of Batesian mimicry and conspicuousness in a sala-mander. Evolution 68(4):1197–1206

Limeri LB, Morehouse NI (2014) Sensory limitations and the maintenance of colour polymorphisms:viewing the ‘alba’ female polymorphism through the visual system of male Colias butterflies. FunctEcol 28(5):1197–1207

Lindstrom L, Alatalo RV, Mappes J (1997) Imperfect Batesian mimicry—the effects of the frequency andthe distastefulness of the model. Proc R Soc Lond B 264(1379):149–153

Llaurens V, Joron M, Thery M (2014) Cryptic differences in colour among Mullerian mimics: how can thevisual capacities of predators and prey shape the evolution of wing colours? J Evol Biol 27(3):531–540

Maia R, Eliason CM, Bitton P-P, Doucet SM, Shawkey MD (2013) pavo: an R package for the analysis,visualization and organization of spectral data. Methods Ecol Evol 4(10):906–913

Marples NM, van Veelen W, Brakefield PM (1994) The relative importance of colour, taste and smell in theprotection of an aposematic insect Coccinella septempunctata. Anim Behav 48(4):967–974

Morrell GM, Turner JRG (1970) Experiments on mimicry: I. The response of wild birds to artificial prey.Behaviour 36(1/2):116–130

Mostler G (1935) Beobachtungen zur frage der wespenmimikry (Observations on the question of waspmimicry). Zoomorphology 29(3):381–454

Odeen A, Hastad O (2003) Complex distribution of avian color vision systems revealed by sequencing theSWS1 opsin from total DNA. Mol Biol Evol 20(6):855–861

Olsson P, Lind O, Kelber A (2015) Bird colour vision: behavioural thresholds reveal receptor noise. J ExpBiol 218(2):184–193

Osorio D, Miklosi A, Gonda Z (1999) Visual ecology and perception of coloration patterns by domesticchicks. Evol Ecol 13(7–8):673–689

Penney HD, Hassall C, Skevington JH, Abbott KR, Sherratt TN (2012) A comparative analysis of theevolution of imperfect mimicry. Nature 483(7390):461–464

R Core Team (2014) R: a language and environment for statistical computing. R Foundation for StatisticalComputing, Vienna

Richards OW (1980) Scolioidea, vespoidea and sphecoidea; hymenoptera, aculeata. Royal EntomologicalSociety of London, London

Richards-Zawacki CL, Yeager J, Bart HPS (2013) No evidence for differential survival or predationbetween sympatric color morphs of an aposematic poison frog. Evol Ecol 27(4):783–795

Rotheray GF, Gilbert F (2011) The natural history of hoverflies. Forrest Text, Cardigan

580 Evol Ecol (2016) 30:567–581

123

Ruxton GD, Sherratt TN, Speed MP (2004) Avoiding attack: the evolutionary ecology of crypsis, warningsignals, and mimicry. Oxford University Press, Oxford

Siddiqi A, Cronin TW, Loew ER, Vorobyev M, Summers K (2004) Interspecific and intraspecific views ofcolor signals in the strawberry poison frog Dendrobates pumilio. J Exp Biol 207(14):2471–2485

Stockman A, Sharpe LT (2000) The spectral sensitivities of the middle- and long-wavelength-sensitivecones derived from measurements in observers of known genotype. Vision Res 40(13):1711–1737

Stubbs AE, Falk SJ (2002) British hoverflies: an illustrated identification guide. British Entomological andNatural History Society, Reading

Svadova K, Exnerova A, Stys P, Landova E, Valenta J, Fucıkova A, Socha R (2009) Role of differentcolours of aposematic insects in learning, memory and generalization of naıve bird predators. AnimBehav 77(2):327–336

Vesely P, Luhanova D, Praskova M, Fuchs R (2013) Generalization of mimics imperfect in colour patterns:the point of view of wild avian predators. Ethology 119(2):138–145

Vorobyev M, Osorio D (1998) Receptor noise as a determinant of colour thresholds. Proc R Soc Lond B265(1394):351–358

Waldbauer G (1988) Asynchrony between Batesian mimics and their models. Am Nat 131:S103–S121Wyszecki G, Stiles WS (2000) Color science: concepts and methods, quantitative data and formulae. Wiley,

New York

Evol Ecol (2016) 30:567–581 581

123